First, unions may improve migrant workers’ employment outcomes by enhancing labor rights protection and regulating working hours. Migrant workers generally occupy a disadvantaged position in segmented urban labor markets, often facing excessive working hours, high labor intensity, and weak enforcement of labor standards (Xue and Li, 2022). According to collective bargaining theory, unions can negotiate with employers to establish more reasonable work-hour arrangements and reduce excessive overtime, thereby improving working conditions (Freeman and Medoff, 1979).

In the absence of unions, firms may extend working hours and intensify labor input to reduce unit labor costs, which can lead to worker exhaustion, higher turnover, and reduced long-term employment stability (Sönmez et al., 2023). By contrast, unions can promote the implementation of standard work-hour systems and limit excessive labor intensity through collective bargaining and workplace supervision (Blanchflower and Bryson, 2025). Reduced working hours not only alleviate physical strain but also contribute to higher labor productivity per unit of time and lower job separation rates (Gihleb et al., 2024). From an employment perspective, the regulation of working hours can stabilize labor supply and improve workers’ ability to remain continuously employed.

Therefore, by protecting basic labor rights and standardizing working hours, unions can enhance employment stability and improve migrant workers’ attachment to the labor market.

Second, unions may affect migrant workers’ employment by promoting the formalization of employment relationships, particularly through increasing labor contract coverage. Employment protection theory emphasizes that formal labor contracts reduce uncertainty by clearly defining rights and obligations for both workers and employers (Ferreiro and Gomez, 2020). In China, many migrant workers remain employed without written contracts, which exposes them to arbitrary dismissal, wage arrears, and unstable employment (Ma et al., 2025).

Unions can encourage and supervise the signing of labor contracts through collective bargaining and rights protection mechanisms (Booth et al., 2022). The existence of labor contracts strengthens job security by reducing informal employment and stabilizing employe–employee relationships. Empirical studies have shown that workers with formal contracts experience lower job turnover and higher employment stability.

For migrant workers, labor contract formalization not only reduces employment uncertainty but also provides an institutional basis for accessing other employment-related benefits, such as social insurance and legal protection (Aizawa et al., 2024). By promoting labor contract signing, unions help migrant workers establish longer-term and more stable employment relationships, thereby improving overall employment outcomes.

Third, unions may improve migrant workers’ employment by expanding access to social security, particularly health insurance, thereby strengthening workers’ labor market attachment. From the perspective of social risk-sharing theory, social insurance reduces workers’ vulnerability to income shocks caused by illness or injury, which in turn affects labor supply decisions and employment continuity (Yixin et al., 2025).

Migrant workers are often excluded from urban social security systems due to informal employment arrangements and employer non-compliance. Unions can play an important role in promoting employers’ fulfillment of social insurance obligations and in raising workers’ awareness of their social security rights (Cheng and Wei, 2024). Health insurance coverage, in particular, reduces the economic burden of illness and lowers the risk of forced labor market exit due to health shocks (Wang et al., 2023).

Improved social security coverage enhances migrant workers’ expectations of future income and reduces perceived employment risk, thereby increasing their willingness to remain employed and stay in urban labor markets (Aizawa et al., 2024). In this sense, unions contribute to employment stability not only through direct labor market interventions but also by strengthening workers’ long-term labor market attachment.

Based on the above theoretical analysis, this study proposes the following hypotheses:

The primary data for this study is derived from the China General Social Survey (CGSS) database and the China Statistical Yearbook. The CGSS database provides nationwide, comprehensive, and continuous micro-level survey data, which offers valuable support for tracking changes in institutional, structural, behavioral, and attitudinal trends, as well as the evolving roles, statuses, and perceptions of social members and groups. Additionally, the database provides rich micro-level data regarding migrant workers’ participation in trade unions. Based on this, the study selects data from the years 2010, 2011, 2012, 2013, 2015, 2017, 2018, and 2021 as the research sample.

Based on the research theme, this study employs a multi-dimensional approach to systematically clean the raw data. First, in consideration of the coverage characteristics of trade union organizations in China and the specificities of the migrant worker group, the sample is restricted to workers who meet all the following criteria: (1) registered as agricultural household (hukou) holders; (2) employed in non-agricultural industries; (3) employed status (excluding employers and self-employed individuals). Second, the sample boundaries are defined according to the legal working age, with male participants aged 16–60 and female participants aged 16–55. Finally, invalid samples are excluded based on the presence of missing key variables (e.g., unreported annual income) or logical anomalies (e.g., negative working hours).

After the standardization process, the final dataset consists of a pooled cross-sectional data with 13,347 observations. The union coverage rate in the sample is 10.64%, a figure comparable to the national union membership rate among migrant workers based on macro-level statistics. From a temporal perspective, the distribution of the sample across the survey years is as follows: 2010 (1,515), 2011 (855), 2012 (1,718), 2013 (1,836), 2015 (1,702), 2017 (2,127), 2018 (2,237), and 2021 (1,357). It should be noted that data for the years 2014, 2016, and 2019–2020 is missing due to the absence of surveys or unpublished data. This intermittent feature will be controlled for in the subsequent analysis using time fixed effects.

To examine the relationship between union participation and migrant workers’ employment, we first estimate a baseline model using Ordinary Least Squares (OLS) regression, which provides an initial assessment of the association while controlling for observable confounders (Equation 1):

In this model, the dependent variable Emp i represents the employment status of migrant workers. union i denotes the union participation variable, indicating whether the individual is a union member. X 1 and X 2 represent individual-level and regional-level control variables for migrant workers, respectively, and ε i is the error term.

However, OLS estimates are likely biased due to endogeneity stemming from self-selection. Union membership is not randomly assigned; it reflects a conscious decision influenced by psychological, social, and contextual factors that are often unobserved. Migrant workers who join unions may differ systematically from non-members in ways not captured by standard controls, such as: (1) Psychological traits: higher risk tolerance, stronger future orientation, greater trust in collective institutions; (2) Social and cognitive resources: broader informal networks, greater awareness of labor rights, enhanced political efficacy; (3) Situational factors: employment in firms with more formalized HR practices or a culture of compliance.

These unobserved characteristics likely influence both the propensity to unionize and employment outcomes, creating a classic case of selection bias. Without addressing this, the estimated “effect “of union membership may merely reflect these underlying differences—a dressed-up correlation rather than a causal relationship.

To address self-selection and omitted variable bias and to identify the causal effect of union participation, we employ a Treatment Effect Model as our primary identification strategy. Following Maddala (1983) for handling binary endogenous regressors, we specify a two-equation system:

In this model, if D ≥ 0 , then union i = 1 otherwise union i = 0 . Z is an independent explanatory variable. Based on the study by Cui et al. (2022), we employ provincial-level data from the China Labor Statistical Yearbook to construct a set of instrumental variables for union involvement. Specifically, these include: (1) the ratio of labor dispute cases mediated by trade unions to the total number of labor disputes in the province; (2) the ratio of union-established training institutions to the total number of union organizations in the province; and (3) the ratio of unions that organize labor competitions to the total number of union organizations. In Equation (3), X 3 represents this vector of instrumental variables, originally used in the selection equation (Equation 2). These variables are hypothesized to influence an individual’s probability of union membership while remaining exogenous to the error term in the outcome equation, thereby addressing potential endogeneity in estimating the effect of union participation on employment outcomes.

To simplify the analysis, we assume that the disturbance terms ( ε i , ζ i ) ~ N ( 0 , 0 , σ 2 , 1 , ρ ) , where ρ represents the correlation coefficient between ε i and ζ i . If ρ ≠ 0 , this indicates the presence of endogeneity in the econometric model. Ω is the inverse Mills ratio, defined as Ω ( c ) = ϕ ( c ) 1 − ϕ ( c ) , where ϕ(c) is the cumulative distribution function of the standard normal distribution. The final estimation equation is as follows:

The treatment effect model relies on the assumption that selection into union membership is jointly determined by observable individual characteristics and exogenous variation in the institutional environment captured by the selection equation. Conceptually, the included covariates proxy migrant workers’ access to information, employment formality, and exposure to collective institutions, which are central channels shaping union participation in China. The provincial-level union activity indicators are intended to reflect institutional capacity and organizational penetration that affect the likelihood of union membership rather than individual employment outcomes directly. We acknowledge that regional labor market institutions may influence employment through multiple channels; therefore, province and year fixed effects, as well as city-level labor market controls, are included to absorb systematic regional heterogeneity. While the exclusion restriction cannot be directly tested, the consistency of results across alternative specifications and matching-based approaches provides supporting evidence that the estimated effects are not solely driven by unobserved selection. Accordingly, we interpret the findings as indicating a plausibly causal employment effect of union participation, rather than definitive causal proof.

Table 2 presents the baseline regression results. Columns (1) through (3) report the estimated associations between union membership and migrant workers’ employment under progressively richer sets of control variables. The results indicate a positive and statistically significant association, providing initial support for Hypothesis H1. However, it is crucial to interpret these OLS estimates as conditional correlations, not as causal effects. The positive coefficient may reflect the influence of unobserved factors, such as workers selecting into unionized firms that offer inherently better employment conditions, or firms with more formalized human resource practices being more likely to host union branches. While we control for a range of individual and regional characteristics, the possibility of omitted variable bias remains.

To credibly identify the causal impact of union membership, this section presents the core results from the Treatment Effect Model (TEM), which is specifically designed to correct for endogeneity. The estimation methods for treatment effects primarily include Maximum Likelihood Estimation (MLE) and Two-Stage Estimation (TS). The regression results for both methods are presented and compared with the OLS regression results, as shown in Table 3.

Table 3 presents the estimation results. Columns (1) through (3) display the results for the relationship between union membership and migrant workers’ employment, estimated using OLS, the two-stage treatment effect model (TEM-TS), and the maximum likelihood treatment effect model (TEM-MLE), respectively. The rejection of the null hypothesis at the 1% level in Column (3) confirms the presence of endogeneity, justifying the use of the treatment effect model. The statistically significant coefficients in Columns (2) and (3) indicate a positive relationship between union membership and employment outcomes after addressing endogeneity concerns. However, these estimates should be interpreted as model-dependent associations rather than definitive structural effects. An important alternative explanation is that union membership may serve as a proxy for employment in larger, more regulated firms that inherently offer better working conditions. While our identification strategy attempts to isolate the effect of union participation, we acknowledge that our estimates may partially reflect this underlying institutional context. Following methodological recommendations (Li and Xu, 2014), subsequent analyses report results based on the more efficient MLE estimator.

Having established a robust causal relationship using the TEM, we proceed to conduct a series of supplementary analyses to assess the sensitivity of our findings to different model specifications and measurement approaches.

While the average treatment effect provides an overall picture, the impact of union membership may vary across different subgroups of migrant workers. To explore this, we conducted subgroup analyses based on skill level, gender, and marital status. The results, presented in Table 5, reveal meaningful heterogeneity in both statistical significance and economic magnitude.

First, regarding labor rights and work-hour regulation, our results show that union membership leads to a significant reduction in weekly working hours (Table 6, Columns 3–4). This aligns with collective bargaining theory (Freeman and Medoff, 1979) and recent findings that unions help curtail excessive overtime, a common issue for migrant workers (Xue and Li, 2022). Reduced work hours alleviate physical and psychological strain, which is critical for sustaining employment. As Sönmez et al. (2023) noted, excessive hours contribute to burnout and higher turnover. By promoting reasonable work schedules, unions enhance job sustainability, thereby directly contributing to employment stability—a core component of our theoretical model.

Second, concerning the formalization of employment relationships, we find a strong positive effect of union membership on the probability of having a written labor contract (Table 6, Columns 5–6). This finding resonates with employment protection theory (Ferreiro and Gomez, 2020) and empirical studies in China documenting unions’ role in improving contract coverage (Booth et al., 2022). A formal contract reduces ambiguity and arbitrariness in employment, providing a legal shield against unjust dismissal and wage arrears (Wang et al., 2019). This institutionalization of the employment relationship is a key mechanism through which unions reduce perceived job insecurity, thereby strengthening migrant workers’ attachment to their current employment and the formal labor market at large.

Third, pertaining to social security and labor market attachment, our analysis confirms that union membership significantly increases the probability of having health insurance (Table 6, Columns 7–8). This extends the findings of Aizawa et al. (2024) and Cheng and Wei (2024), highlighting unions’ role in bridging the social protection gap for migrant workers. Health insurance mitigates the financial shock of illness, reducing the risk of involuntary unemployment due to health crises (Wang et al., 2023). By lowering the economic vulnerability associated with health risks, unions enhance migrant workers’ long-term employability and their willingness to remain in urban labor markets. This mechanism underpins the broader concept of urban integration and labor market attachment theorized in Section 2.3.

Our mechanism analysis demonstrates that the employment effect of unions is not monolithic but operates through a synergistic combination of institutional formalization (contracts), risk mitigation (social insurance), and well-being enhancement (regulated hours). These empirical results collectively substantiate the theoretical proposition that unions function as vital institutional supports that reduce employment uncertainty.

Furthermore, our findings on labor contracts and social insurance provide empirical grounding for the urban integration pathway. Secure employment relations and social safety nets are foundational for migrant workers’ long-term urban settlement and social inclusion (Wang and Shen, 2022). While our primary focus and variable selection in Table 6 centered on proximal labor market outcomes (income, hours, contracts, insurance), these very factors are established precursors to broader socio-economic integration. Future research could incorporate direct measures of urban settlement intention or social network expansion to explicitly test this extended pathway.

The positive effects documented here are consistent with prior studies on unions’ protective roles in China (Ma, 2024; Ma and Komatsu, 2024) but extend the literature by explicitly linking these protective functions to employment stability outcomes rather than solely to wage levels. This addresses a gap noted by Sayre (2023), who called for more research on how institutional buffers affect employment-related insecurity. Our results suggest that in the Chinese context, unions partially offset the institutional disadvantages stemming from the hukou system by fostering more stable and protected employment conditions.

While this study provides robust evidence supporting the positive association between union participation and migrant workers’ employment, several limitations should be acknowledged. First, despite our use of a treatment effect model and instrumental variables, endogeneity due to self-selection and unobserved heterogeneity cannot be fully ruled out. Union membership may still serve as a proxy for employment in more formalized sectors with better labor conditions. Second, our data are cross-sectional and lack longitudinal tracking, which limits our ability to examine the dynamic effects of union participation on employment stability over time. Third, the binary measure of union membership does not capture variations in union activity intensity or functional effectiveness, which may influence employment outcomes differently. Fourth, our mechanisms are inferred from observable outcomes (e.g., contracts, insurance) but lack direct measures of psychological constructs such as perceived job security or belongingness.

Future research could address these gaps by: (1) utilizing panel data or natural experiments (e.g., policy shocks in unionization campaigns) to strengthen causal identification; (2) incorporating mixed methods, including qualitative interviews, to explore how union participation shapes migrant workers’ psychological well-being and social integration; (3) developing more nuanced measures of union engagement and workplace institutionalization; and (4) examining the long-term effects of union participation on career progression and urban settlement.

Given the significant benefits identified, policymakers should prioritize expanding unionization in private enterprises, small and medium-sized firms, and labor-intensive industries where migrant workers are concentrated and formal protection is often weakest. Concrete actions could include: (a) simplifying union registration procedures for enterprises with a high proportion of migrant workers; (b) offering tax incentives or regulatory compliance benefits to firms that achieve high union membership rates; and (c) launching targeted campaigns to recruit female, low-skilled, and informal workers into union membership.

To operationalize the mechanism of employment formalization, we recommend establishing a national digital labor contract and rights-monitoring platform integrated with social insurance systems. This platform should: (a) enable real-time, mandatory registration of all labor contracts, with automatic alerts for non-compliance or contract expiration; (b) provide a direct channel for workers to report violations of working-hour regulations or wage arrears, linked to union and labor inspection offices; and (c) publicly disclose enterprise compliance ratings to strengthen market and social supervision.

To address the social security gap, unions should be empowered to administer or facilitate portable, subsidized social insurance packages tailored for migrant workers. Specific measures include: (a) piloting “union-managed health insurance pools” that allow seamless coverage across jobs and regions, with premium subsidies for low-income members; (b) mandating unions’ role in verifying and enrolling workers in existing urban insurance schemes; and (c) using union halls or mobile stations for on-site insurance enrollment and claims assistance.

Building on our finding that unions reduce excessive hours, regulatory bodies should collaborate with unions to develop and enforce industry-specific working hour benchmarks. This involves: (a) tripartite negotiations (government, union, employer associations) to set realistic hourly limits per sector; (b) requiring unions to appoint “working-hour monitors” in large workplaces; and (c) introducing a mandatory rest-period log for high-intensity jobs, subject to random union audits.

To ensure unions can effectively deliver the mechanisms identified, state funding should support: (a) specialized training programs for union representatives on labor law, contract negotiation, and mental health support; (b) grants for unions to provide direct services such as legal aid clinics, vocational upskilling courses, and counseling hubs; and (c) performance-based funding tied to measurable outcomes in contract coverage, insurance enrollment, and grievance resolution for migrant workers.

These recommendations are directly derived from our empirical findings and are designed to translate the protective functions of unions into scalable, institutionalized practices. Future policy evaluations should assess the impact of these interventions not only on employment stability but also on psychological well-being and long-term urban integration of migrant workers.